VisualiseCAM on models with multiple input heads

[koriavinash1]

I'm trying to run gradcam analysis on the network with multiple input heads, but facing errors like:

    seed_input = X2)
  File "/home/brats/miniconda3/envs/dl/lib/python3.6/site-packages/vis/visualization/saliency.py", line 245, in visualize_cam
    return visualize_cam_with_losses(model.input, losses, seed_input, penultimate_layer, grad_modifier)
  File "/home/brats/miniconda3/envs/dl/lib/python3.6/site-packages/vis/visualization/saliency.py", line 166, in visualize_cam_with_losses
    _, grads, penultimate_output_value = opt.minimize(seed_input, max_iter=1, grad_modifier=grad_modifier, verbose=False)
  File "/home/brats/miniconda3/envs/dl/lib/python3.6/site-packages/vis/optimizer.py", line 132, in minimize
    seed_input = self._get_seed_input(seed_input)
  File "/home/brats/miniconda3/envs/dl/lib/python3.6/site-packages/vis/optimizer.py", line 94, in _get_seed_input
    desired_shape = (1, ) + K.int_shape(self.input_tensor)[1:]
  File "/home/brats/miniconda3/envs/dl/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py", line 593, in int_shape
    return tuple(x.get_shape().as_list())
AttributeError: 'list' object has no attribute 'get_shape'

any comments on this can be fixed?

Thank you